/CAPTCHA-Breaker

This repository contains the code for generating dataset of different variations of text based CATPCHAs which are further sent to Pix2Pix (GAN) for generating more CAPTCHAs of the selected variation and then passed to a CNN which recognizes the characters in the CAPTCHA.

Primary LanguagePython

CAPTCHA-Breaker

This repository contains the pipeline for breaking a text-based CAPTCHA using python and deep learning. The pipeline has the following 3 modules:

  1. CAPTCHA-Generator is used to generate a limited dataset of text-based CAPTCHAs and there are different variations to choose from.
  2. Pix2Pix is a GAN (Generative Adversarial Network) which is used to generate more of the selected variation of text-based CAPTCHAs.
  3. CAPTCHA-Classifier is a CNN (Convolutional Neural Network) which is used to recognize the characters contained in the text-based CAPTCHA.

Prerequisites

  • Tensorflow 1.4.1
  • Python 3.x
  • OpenCV

How to run

  • Download the datasets and extract them inside the CAPTCHA-Generator\Images directory to be able to generate different variety of text-based CAPTCHAs having different background and textured alphabets.
  • Run the CAPTCHA-Breaker.bat file and you are good to go.